Diagnostic tests in ado file

[Martijn 9000]

Dear Stataexperts,

I have a question about postestimations of the "cluster2" ado file. I have estimated a model using two clusters (command: cluster2 dv iv1 iv2 iv3 iv4, fcluster(var) tcluster(var)). However, by using this model (and this ado file) I cannot test for model assumptions. For example, when I type in "estat hettest" or "estat imtest", stata says: "subcommand estat imtest is unrecognized".

I am able to use for example "estat ic". Besides, it is possible to use "estat hettest" if I would have just a normal regression as a model (without the ado file: reg dv iv1 iv2 iv3 iv4).

Does anyone know what I can do to solve this issue? I have tried (almost) everything!!

[Nick Cox]

Please use a real name as requested in the FAQ,

[Martijn 9000]

I cannot change it anymore.. And I only need an answer on this (hopefully) small question

[William Lisowski]

Unless you're particularly fortunate and someone here already familiar with the cluster2 command sees and responds to your question, you're unlikely to find help, since the Stata command search cluster2 yields no results for the curious and potentially helpful readers to investigate. So this is far from a small question. A quick guess, and it's nothing more than that, is that the apparently user-written cluster2 command does not support the postestimation subcommands you want to use. Try issuing the command help cluster2 and seeing what it says about postestimation with cluster2.

You might benefit from reviewing the Statalist FAQ linked to from the top of each page, paying particular attention to the information in item 6 on changing your registered name, and in items 9-12 on posing questions in a way that maximizes the chances a reader can assist you, especially (in this case) the advice that you state the source of user-written commands.

[Jorge Eduardo Perez Perez]

Estimation with cluster robust standard errors are robust to heteroskedasticity and correlation within the cluster, so it doesn't make sense to test for heteroskedasticity here, since you assumed that the error terms are heteroskedastic for your estimation.

This should be a FAQ, it arises frequently.